search

BIDS-projects / etl

Extraction, transformation, and loading for topic modeling
0 stars 0 forks source link

Figure out where the rest of the boilerplates are coming from #5

Open don-han opened 8 years ago

don-han commented 8 years ago

https://github.com/BIDS-projects/topic-modeling/issues/21#issuecomment-188669165

chewisinho commented 8 years ago

I found the location of the topic words "email address", "spambots", and "JavaScript enabled". They all come from one source: www.sdav-scidac.org

The problem is not with the scraper (we are not getting blocked) and not with the HTML parser - the problem is the website itself.

I have a super simple fix that actually filters out the words really effectively. Just look for this sentence:

"This email address is being protected from spambots. You need JavaScript enabled to view it."

Removing the above sentence from the corpus produced pretty good results.

Things to consider: 1) What other boilerplates need to be removed? 2) How should we remove these boilerplates? Should we hard code the fix (because it's astonishingly effective), or should we try adding these words to the stop words, etc.? Thoughts?